Field of the Invention
The present invention relates to a communications network and a method of operating a communications network.
Description of the Related Art
Recently, the demand for streaming video to a computer via the Internet has grown strongly. This has led to a need to supply increasing amounts of video material over local communication networks (including the copper pairs used by telephone network operators or the coaxial cables used by cable television network operators).
In telephony networks this additional demand is being met partly by the introduction of Digital Subscriber Loop (DSL) technology. As its name suggests, this technology carries digital signals over the local copper loop between a user's home and a local telephone exchange. Data-rates of several megabits per second to the user's home have become commonplace. Advances in this technology now lead to much higher rates, 20 Mbit/s and above and with plans in the industry to offer 50 Mbit/s and above to a substantial proportion of broadband end users in the relatively near future. Using DSL, the digital signal is conveyed between modems placed at either end of the copper loop. The advantages of statistical multiplexing have led to the digital signals being organized into transport packets (whether they be Asynchronous Transfer Mode (ATM) packets or Ethernet packets. Over either of these are carried Internet Protocol (IP) packets conveying all of the broadband services.
Cable networks have also been upgraded to carry broadband services to user's homes. Substantial numbers of users currently receive broadband services over cable and, again, the services are conveyed using IP packets over Ethernet.
Video material requires a data rate which varies between 1.5 Mbps (for a quality comparable to that offered by a video cassette recording) to 20 Mbps (High-Definition Television). In DSL or cable networks, these higher application data rates means that a mechanism is needed to manage contention for the capacity available towards the user's home. This capacity management includes both the maximum available capacity for any one user and the shared capacity towards the DSL Access Multiplexer (DSLAM) or Cable Modem Termination System (CMTS). This capacity (sometimes termed “backhaul” capacity) may be shared by several hundred users, and contention for the capacity will also need to be managed as users demand more choice in the material they view.
The backhaul capacity may be divided into VLANs, where each VLAN is the aggregate bandwidth that a single internet service provider can exploit to deliver services to many end-users. In this case, the active users who are receiving internet services via this service provider will share the capacity of this VLAN aggregate. In an alternative arrangement, several internet service providers may jointly share the backhaul capacity and some means may be provided so that each service provider obtains a fair share.
Under the first arrangement, where the capacity of a VLAN is available to a single service provider, there is freedom for that service provider to make unique choices about cost and QoS. The choice is a balance between insufficient capacity would result in frequent congestion and more dissatisfaction among the end users. Too much capacity results in excess cost for the provision of services.
Under the second arrangement, where several service providers share capacity, there may be more opportunity to exploit unused capacity that belongs to one of the fair shares. This has advantages but also some disadvantages. The advantage is that the cost of aggregate bandwidth may be kept lower for each of the service providers. Instead they may rely on the availability of unused bandwidth when otherwise there would have been dissatisfaction among their end users due to congestion. The disadvantage is that the availability of extra bandwidth is uncertain, being dependent on the patterns of usage of end users of each of the sharing service providers. Another disadvantage is the non-uniqueness of the cost/QoS offering of each of the service providers. However, this disadvantage can be overcome by altering the nature of the fair-share controls of the aggregate bandwidth, so that one service provider is preferred over another in terms of the amount of capacity its users are entitled to get when loading levels are high.
An arrangement where capacity is shared among several service providers and the capacity is unequally divided among them when loading levels are high can lead to even greater uncertainty that extra capacity is there when needed, especially for those service providers that receive a lower amount of capacity that the average.
It is the management of this issue to remove the uncertainty that extra bandwidth is there when needed and to provide just enough bandwidth, on backhaul links or other network links, that is the subject of the present patent.
In a conventional circuit-switched telephone network, the problem of contention for scarce telecommunication resources is dealt with by simply preventing a user from receiving (or sending) any traffic unless the necessary capacity to carry that traffic can be reserved beforehand. The capabilities of multi-services packet networks have also been developed to include call admission control schemes—examples include the Resource Reservation Protocol (RSVP). Although such schemes can prevent congestion when all new communications or calls are admitted or rejected using these principles, QoS management must also manage so-called “elastic” traffic where there is potentially a need for a minimum guaranteed rate but frequently a desire to transmit the flow as fast as possible, subject to network congestion constraints and constraints on maximum sending rates.
An alternative to the use of connection admission control in packet networks is to use reactive flow control. These schemes allow users access to communications resources but attempt to cause senders to decrease their sending rate on the onset of congestion. The scheme used for reliable transmission across the Internet (Transmission Control Protocol) is the most common example. This is unsuitable for video flows however, since real-time video servers cannot reduce their sending rate.
Most flow control schemes, applied to elastic applications, result in some reduction of the rate available at the onset of congestion. Some flow control schemes are more sophisticated, classifying traffic into different classes, with some classes being more likely to suffer packet delay or discard than others. In situations where such classification is not available or where most traffic is within one class, alternative solutions must be provided. One such alternative solution which concentrates the adverse effects of ATM cell discard on one IP packet at a time is described in ‘Early Selective Packet Discard for Alternating Resource Access of TCP over ATM-UBR’ by Kangsik Cheon and Shivendra S. Panwar, in the Proceedings of IEEE Conference on Local Computer Networks LCN 97, Minneapolis, Minn., Nov. 2-5, 1997, incorporated by reference herein.
A discussion toward the Internet based Next Generation Network (NGN) is actively progressed around the standards bodies including the ITU-T (International Telecommunication Union-Telecommunication Standardization Sector) and ETSI (European Telecommunications Standards Institute), IEEE (Institute of Electrical and Electronics Engineers), IETF (Internet Engineering Task Force), and etc. The roles of the standard bodies are different. The IEEE and IETF develop the core technology for specific problems in layer 2 and layer 3, respectively. ITU-T and ETSI develop the network architecture and control procedure.
A QoS control or resource control architecture has been developed in the several standard bodies. To name a few, they are ITU-T, ETSI, Cable Lab, 3GPP, MSF, and the DSL forum. Among those organization, CableLab, and DSL forum, 3GPP, and ETSI define the QoS control architecture in a particular case while ITU-T defines the generic architecture that can cover the outcomes of other standard bodies.
CableLab defines the dynamic QoS (DQoS) control architecture as described in PacketCable Specification “PacketCable Dynamic Quality-of-Service” for the Hybrid Fiber and Coaxial (HFC) network. The control architecture is designed for the uniqueness of the HFC network. In the HFC network, multiple CMs (Cable Modems) share an upstream channel to CMTS (Cable Modem Termination System). The bandwidth is controlled based on layer 2 MAC protocol called DOCSIS (Data Over Cable System Interface Specification) as described in “Data-Over-Cable Service Interface Specifications, Radio Frequency Interface Specification”, by Cable Television Laboratories, Inc., Jul. 30, 2003. The layer 2 level QoS guarantee mechanism is defined from the DOCSIS version 1.1. The goal of the DQoS is supporting the QoS guaranteed service through HFC network.
DQoS defines the procedure of the call setup signaling and the dynamic QoS control on DOCSIS interface. In the architecture, the CMS (Call Management Server) I Gate controller controls the call establishment. The guaranteed bandwidth between CM and CMTS is reserved dynamically during the call setup signaling. The CMS/Gate Controller triggers the layer 2 or layer 3 QoS signaling to reserve the bandwidth in the HFC network by sending commands to CM, CMS, or MTA (Multimedia Terminal Adapter).
DQoS has been refined through version 1.0, 1.5, and 2.0. Version 1.0 defines the basic call setup signaling procedure for both embedded MTA and standalone MTA. The embedded MTA can initiate the dynamic layer 2 QoS signaling while a standalone MTA initiates IP level QoS signaling. Version 1.5 and 2.0 defines the QoS control architecture when SIP (Session Initiation Protocol) based call setup signaling is used. DQoS 2.0 is defined especially for interoperability with IP Multimedia Subsystem (IMS) which is the SIP based call setup architecture developed in 3rd Generation Partnership Project (3GPP). PacketCable Multimedia, as described in PacketCable Specification “Multimedia Specification”, Dec. 21, 2005, has been developed for simple and reliable control for the multimedia service over cable network. It defines the service delivery framework for the policy based control on multimedia service. The simple procedure for time or volume based resource authorization, resource auditing mechanism, and security of the infrastructure are defined in PacketCable Multimedia.
Such developments as this strongly suggest that new QoS mechanisms should take account of, and build on top of, the underlying deployment of QoS controls.
Again, the DSL forum defines the resource control at the DSL (Digital Subscriber Line) access network, as described in Technical Report 59 DSL Forum “DSL Evolution—Architecture Requirements for the Support of QoS-Enabled IP Services”. Unlike Cable network, DSL modem is connected to the subscriber through the dedicated line. Layer 2 level dynamic QoS control between DSL modem and Digital Subscriber Line Access Multiplexer (DSLAM) is not required. The DSL forum focuses more on resource control in the home network especially resource control of multiple terminals behind the home gateway.
The resource control architectures defined in the above mentioned two standard bodies—PacketCable and DSL Forum focus on a specific transport technology (i.e., HFC network and DSL network). The scope of DQoS and DSL forum is mainly within network operator's view. Unlike these, RACF (Resource and Admission Control Functions), as described in ITU-T recommendation Y.2111 “Resource and Admission Control Functions in NGN”, and RAGS (Resource and Admission Control Sub-system), as described in ETSI ES 282 003 V1.1.1 (2006-03), “Resource and Admission Control Sub-system (RAGS); Functional Architecture”, define the resource control architecture in more general aspect.
The QoS control architecture in both RACF and RAGS are closely related with 3GPP (3rd Generation Partnership Project) effort. The 3GPP is originally founded for developing new service architecture over cellular network, especially for GSM (Global System for Mobile communication) network. During this effort, 3GPP developed the IMS (IP Multimedia Subsystem) for controlling the IP multimedia services in the areas of session control, service control, and management of database of the subscribers. Even though IMS is initially developed for the evolution of GSM cellular network, its framework can be applicable for any types of transport technologies. The IMS architecture has been adopted to the other QoS control architectures such as 3GPP2 MMD (Multimedia Domain), ETSI TISPAN (Telecoms & Internet converged Services & Protocols for Advanced Networks), and ITU-T NGN. Thus, both RAGS and RACF are interoperable with IMS.
In general, RACF and RAGS are very similar with each other. The two standards bodies are closely interacted in developing their architecture. There is no significant conflict between the two, but there are still differences, as described in ITU-T NGN-GSI Contribution, “Comparison of TISPAN RAGS and ITU-T RACF”. One of differences is the range of the control region. The control region of RAGS covers the access network and the edge of the core network. The access network is defined as the region where the traffic is aggregated or distributed without dynamic routing. The resource control in the access network is done in layer 2 level. The core network is the region that the IP routing starts. The core network is out of scope in the RAGS. RACF, however, covers both core and access network. RACF covers both fixed and mobile networks while RAGS is defined for the fixed network. For the control mechanism, the RACF defines more control scenarios than RAGS. Therefore, RAGS is considered as a subset of RACF.
ITU-T defines QoS control functions based on its NGN architecture. One of the important concepts in the ITU-T NGN architecture is the independence of the transport and the service, as described in ITU-T recommendation Y.2012 “Functional Requirements and Architecture of the NGN”. The transport is concerning about the delivery of packets of any kind generically, while the services are concerns about the packet payloads, which may be part of the user, control, or management plane. In this design principle, the NGN architecture is divided into two stratums—Service Stratum and Transport Stratum. Under the concept of the independence of a service and transport functions, the network resource and reliability are guaranteed by the network side upon request from the service stratum. Service Stratum is responsible for the application signaling and Transport Stratum is responsible for reliable data packet forwarding and traffic control. The service stratum can be a simple application server or a full-blown system such as IMS (IP Multimedia Sub-system).
Transport control function is located in Transport stratum interfacing with the Service stratum. It determines the admission of the requested service based on the network policy and the resource availability. It also controls the network element to allocate the resource once it is accepted. Resource and Admission Control Functions (RACF) is responsible for the major part of the admission decision and resource control of the transport function. Details of RACF mechanism can be found in “Overview of ITU-T NGN QoS control”, by Jongtae Song, Mi Young Chang, Soon Seok Lee, and Jinoo Joung, IEEE Communication Magazine, Vol. 45, No. 9, September 2007 and ITU-T recommendation Y.2111 “Resource and Admission Control Functions in NGN”, incorporated by reference herein.
This developing infrastructure needs to be taken account of when considering new QoS mechanisms.
Review of current per-flow QoS controls Flow level transport technology is not a new concept. The core technologies for traffic management schemes such as flow level scheduling, policing, and sharing are already available in a commercial product, as described in “Flow based control for Future Internet” by Jongtae Song, presented in Future Internet Forum (FIW) in July 2007, incorporated by reference herein. The current deployment of flow base control, however, is limited only at the edge of the network. Typical examples of flow base control are traffic monitoring and packet inspection, PacketCable access, session border controller, edge router, and interworking between two networks. They are mostly stand-alone solution at the edge of the network.
However, flow level traffic control only at the edge cannot guarantee the flow level QoS. Furthermore, DiffServ guarantees the QoS only if the premium traffic load is very low (˜under 10%), as described in “Providing guaranteed services without per flow management” by I. Stoika and H. Zhang in CM SIGCOMM, September 1999, pp. 91-94, incorporated by reference herein. On the other hand, having scalable control architecture for flow level traffic control along the data path is a challenging issue, because the number of flows in a network is huge.
There are several schemes proposed for the scalable control of traffic using flow level mechanisms. These are listed below.
(1). Flow Aware Network (FAN) France Telecom proposed a Flow Aware Network (FAN), as described in “A new direction for quality of service: Flow-aware networking” by S. Oueslati and J. Roberts in Proc. Conference on Next Generation Internet Network (NGI), April 2005, incorporated by reference herein. FAN applies three different regimes based on the network status. They are the “transparent regime”, “elastic regime”, and “overload regime”. The transparent regime is applied when the network has no congestion at all. The elastic regime is applied when the network experiences the occasional traffic congestion because of a few high rate data flows. The overload regime is applied when the traffic overloads the link capacity in the network.
No traffic control is required in the transparent regime. The traffic control is effective only in the overload or elastic regime. In the elastic regime, the network enforces the bandwidth limit for every flow. Every flow is assigned the same amount of bandwidth. In the overload regime, new flows are blocked to protect existing flows. To reduce the control complexity, an implicit approach is preferred where no signaling is required for controlling the network. Each node makes locally optimal decision based on local observation.
The main focus of FAN is the simplicity. It requires no signaling. Only implicit admission control is required upon congestion. Although the control mechanism is very simple, it is shown that the network is stabilized remarkably in FAN. However, this architecture is designed mainly for network stabilization aspect. Every flow is treated equally. In order to support various of QoS requirement for individual flow, this architecture should be improved.
(2). Flow Sate Aware (FSA) technologies FSA is developed to provide different QoS for the individual flow. FSA defines the service types based on typical example of Internet services, as described in ITU-T Recommendation Y.2121, “Requirements for the support of stateful flow-aware transport technology in an NGN” and “Changing the internet to support real-time content supply from a large fraction of broadband residential users” by J. L. Adams, L. G. Roberts, and A. I. Jsselmuiden, BT Technology Journal, Vol. 23, No. 2, pp 217-231, April 2005, incorporated by reference herein. They are Maximum Rate (MR), Guaranteed Rate (GR), Variable Rate (VR), and Available Rate (AR). GR is designed for applications requiring guaranteed bandwidth for the entire duration of the flow. MR is designed for streaming media such as video and voice. AR is designed for data traffic flow where the application can setup the flow rate at the maximum rate that the network can currently support. VR is the combination of AR and MR. VR could be used for obtaining a maximum response time for a transaction (e.g., a stock trade with maximum transaction time). The MR portion guarantees the minimum guaranteed bandwidth and AR portion is for use available network resource. FSA divides the network resource into two portions. One is Fixed Rate (FR) and the other is Network Rate (NR). FR is requested when flow needs a fixed rate available during the service. NR is requested when flow sends buffered data using network available bandwidth. Service type GR and MR request FR, AR requests NR, and VR requests both NR and FR. The detail requirement is defined in ITU-T Recommendation Y.2121, “Requirements for the support of stateful flow-aware transport technology in an NGN”.
FR and NR are requested by the signaling, as described in ITU-T Study Group 11, Draft Recommendation Q.flowstatesig on signaling protocols and procedures relating to Flow State Aware access QoS control in an NGN, Editor J. L. Adams, incorporated by reference herein, and every node along the path configures its resource based on the requested FR and NR. For the call setup signaling, the source node and destination node exchanges the control messages. [FIG. 1] describes the signaling procedure for the service type MR, GR, and AR. In the ingress FSA (iFSA) and egress FSA (eFSA) exchanges the request, response, confirm, renegotiate, and confirm message for request the transport resource. For MR, iFSA sends the data traffic before receiving the response from eFSA. MR is designed based on the concept of the conditional guaranteed bandwidth, as described in “Changing the internet to support real-time content supply from a large fraction of broadband residential users” by J. L. Adams, L. G. Roberts, and A. I. Jsselmuiden, BT Technology Journal, Vol. 23, No. 2, pp 217-231, April 2005, and ITU-T Recommendation Y.1221 Amendment 2 (2005), Traffic control and congestion control in IP-based networks, incorporated by reference herein. For GR, it needs to know the explicit start and ending time of the flow. Therefore, it sends confirm and close messages for acknowledging every transit node reserves and release the requested bandwidth. The service type AR is designed to use network available resource. iFSA and other FSA nodes continuously monitor the network available resource and adjust the NR accordingly.
Both approaches, FAN and FAS, give an insight for flow based traffic control. FAN shows that even very simple flow level traffic control can stabilize the network efficiently. FSA shows that the network resource can be divided into FR and NR. It also indicates that the transit nodes should be controlled for end-to-end flow level QoS.
However, the two approaches have outstanding issues. As mentioned earlier, FAN is not designed for supporting various QoS requirement of the service. Its main objective is stabilizing the overall network performance. In this viewpoint, FAN treats every flow equally. This may stabilize the transport network in general, but the network provider cannot generate additional profit, because FAN cannot support the service that has special QoS treatment. Good business model is hardly found in this case.
FSA is designed for supporting various QoS requirement. Its implementation can be done in both the in-band signaling, as described in ITU-T Study Group 11, Draft Recommendation Q.flowstatesig on signaling protocols and procedures relating to Flow State Aware access QoS control in an NGN, and out-of-band signaling. The in-band signaling procedure requires the every node exchanges the request and response. The request need to be examined by the all the transit node. The destination node generates the response message, and source node finds the agreed rate from the response message. In this approach every FSA node should maintain the flow state.
Requiring FSA signaling feature in every user terminal is possible. However, by making the terminal independent of FSA, we can have several benefits. First, the terminal usually has different capability. The network architecture should be flexible enough to support multiple types of terminal in a network. The terminal can support transport QoS signaling but also has application signaling. The application signaling is common for all terminal types. In order to support more terminal types, the QoS signaling of terminal should be designed in application level. Second, the network security is important in managed network. Enabling the signaling function in the terminal may cause the security hole in the network. For resolving this problem, ITU-T Recommendation Y.2121, “Requirements for the support of stateful flow-aware transport technology in an NGN” specifies the mechanism to authorize the in-band signaling in the application signaling phase. FSA signaling initiated in the network side from the network edge can be another option to avoid the security problem.
In both FAN and FSA approaches, focus is mainly in the transport control. In order to take account of the existing deployment of QoS functions, the concept of RACF function needs to be considered.
(3). FSA with out-band signaling. In this proposal, the FSA signaling is combined with RACF. CPE (Customer Premises Equipment) or user terminal should be able to request the flow level resource in any kind of application signaling. In this aspect, the CPE and user terminal should be protocol independent.
Second, this proposal focuses flow-based control in the access network, not the core. In the access network, user data traffic is statically routed to the edge of the core network, and the downstream data traffic is statically forwarded from the edge of the core network to the end user. Core supports both IP based dynamic routing and layer 2 based static forwarding. The traffic volume, number of flows, and dynamicity of traffic are different in the core and access. Traditional access network controls the bandwidth based on subscribed bandwidth per user in L2 level. For flow level traffic control, however, the bandwidth should be controlled by individual flow. Flow awareness capability is required in the access nodes. The static packet forwarding and scheduling in the flow level granularity is required in the access network. The call by call traffic control and policy enforcement from control plane (e.g., RACF) should be done in micro flow level. In the core, the number of flows is high and call by call flow level control in RACF is difficult to achieve. In the core side network, therefore, the traffic should be controlled in aggregate level rather than micro flow level. The reliability and monitoring capability will be more important in the core. The flow based traffic control and the aggregated traffic control should be translated at the edge of the core network.
These design principles are further illustrated in FIG. 2.
However, none of the above proposals provides a method of managing contention in a packet network which allows flow-based QoS mechanisms to operate without end-user signaling and support:                Preference priority control of some flows in the event of congestion or sudden re-routing of traffic in the event of a network link failure.        Admission of variable rate, delay-sensitive flows requiring some minimum guaranteed bandwidth.        Management of the fastest transfer time (highest available transfer rate).        
In ‘Flow State Aware QoS Management Without User Signaling’, U.S. Provisional Application No. 61/118,964 filed Dec. 1, 2008, and ‘Flow State Aware QoS Aggregate Management Without User Signaling’, U.S. Provisional Application No. 61/185,843 filed Jun. 10, 2009, solutions are described that do not require signaling. Both solutions assume a fixed assignment of capacity is provided to each aggregate VLAN. Solution in U.S. Provisional Application No. 61/118,964 filed Dec. 1, 2008 gives every flow either a guaranteed rate or some assignment of a minimum rate and some assignment of the remaining unused capacity which is adjusted according to the number of flows and the preference priority of a flow. Solution in U.S. Provisional Application No. 61/185,843 filed Jun. 10, 2009 does not assign capacity to a single flow and does not discard any packets of any flows until the fixed capacity of an aggregate is nearing congestion. It causes less packet deletions than U.S. Provisional Application No. 61/118,964 filed Dec. 1, 2008, although, if there is evolution towards some use of signaling for some of the services and for some of the flows, U.S. Provisional Application No. 61/185,843 filed Jun. 10, 2009 is potentially more unfair on those flows that are being policed through signaling while other flows, established without signaling, are free to change their rates arbitrarily. In this case, U.S. Provisional Application No. 61/118,964 filed Dec. 1, 2008 is believed to provide a fairer arrangement. The aforesaid patent applications are incorporated by reference herein.
However, both solutions in U.S. Provisional Application No. 61/118,964 filed Dec. 1, 2008 and U.S. Provisional Application No. 61/185,843 filed Jun. 10, 2009 assume the aggregate VLAN capacities are constant. The choice of the capacity is important to the service provider. There is a need to avoid congestion becoming too frequent. However, there may be uncertainty about the traffic load, especially with the take-up of new applications among the end users. Therefore, it is typical for the service provider to provide excess capacity to ensure congestion is infrequent. But, as application data rates continue to increase, this approach of always having excess capacity becomes more and more costly for the service provider and, potentially, more costly for the user.
Thus, novel methods of operating a communications network are needed.